Better Handling of 'None' in ARG Drug Table

CHANGELOG.md

@@ -1,5 +1,6 @@
 # Unreleased
 
+* Fixed an issue where the string "None" in the drug table would be parsed differently by different versions of pandas (#175).
 * Upgraded to pandas version 2.
 * Added CGE-predicted phenotypes to Pointfinder output.
 * The resfinder.tsv and pointfinder.tsv outputs now contain a Notes column.

staramr/databases/resistance/ARGDrugTable.py

@@ -24,7 +24,9 @@ def __init__(self, file=None, info_file=DEFAULT_INFO_FILE):
         self._file = file
 
         if file is not None:
-            self._data = pd.read_csv(file, sep='\t', dtype=self.DTYPES)
+            # "None" is recognized as a NA/NaN string in pandas 2.
+            # However, in pandas < 2, "None" is not a default NA value, so we must be explicit.
+            self._data = pd.read_csv(file, sep='\t', dtype=self.DTYPES, na_values="None")
 
     def get_resistance_table_info(self):
         """
@@ -41,4 +43,10 @@ def _drug_string_to_correct_separators(self, drug):
         :param drug: The drug string.
         :return: The drug string with correct separators/spacing.
         """
-        return ', '.join(drug.split(','))
+
+        if type(drug) is str:
+            result = ', '.join(drug.split(','))
+        else:
+            result = drug
+
+        return result

staramr/tests/unit/results/test_ARGDrugTablePointfinder.py

@@ -0,0 +1,20 @@
+import logging
+import unittest
+import pandas
+
+from staramr.databases.resistance.pointfinder.ARGDrugTablePointfinder import ARGDrugTablePointfinder
+
+logger = logging.getLogger('ARGDrugTablePointfinderTest')
+
+
+class ARGDrugTablePointfinderTest(unittest.TestCase):
+
+    def setUp(self):
+        self.arg_drug_table = ARGDrugTablePointfinder()
+
+    def testNoneEntry(self):
+        # Tests when the entry for the drug is "None"
+        drug = self.arg_drug_table.get_drug("escherichia_coli", "parC", 57)
+
+        # Specifically, we're interested in it not crashing.
+        self.assertTrue(pandas.isna(drug))